OpenClaw, but for enterprise.
Turn into an AI employee you can trust.
OpenFigs is the open skeleton for one back-office AI employee — plain files and prompts you read end to end. You clone it, you don't install it.
Ledger's operating guide — who it is and the loop it runs every close. You read it end to end before it ever runs.
# Ledger — operating guide You are an AI accountant: you own the monthly close, do it, learn from each close, and report to your controller through Figs. ## The loop (every run) Orient → Reconcile → Record → Report
01Why you can trust it
Thin on purpose.
A back office can't sign off on a super-agent it can't see. So OpenFigs is deliberately small: one job, plain files, every line readable before it runs. Nothing to install, and nothing to take on faith.
See the files yourself02Your team's view
One push, and they're on the org chart.
Each employee reports through the open .figs protocol. Run figs push and it shows up in Figs — an org chart your whole team can see, with the handoffs that need a human. Run this once per job; they all land in the same place.
See it at figs.so
FAQ
Questions & answers.
A folder of files. It turns Claude Code or Codex into one AI employee with a real job — say, an accountant that closes your books. You read every file. Nothing to install, nothing hidden.
OpenClaw is one big agent you install and can't see into. OpenFigs is one small agent that's just files you can read. Same idea, made safe for a company.
No. It works on its own. When you want your team to see what it's doing, run figs push — it shows up on the org chart at figs.so.
Nothing. It's open source (MIT). You clone it, you own it.